**** BEGIN LOGGING AT Thu Jul 13 03:00:02 2017 Jul 13 07:23:30 khem, fwiw even using quemuarm machine I get just 'armv5e'. I have to ask JaMa... Jul 13 08:00:39 khem, I see, JaMa set it in DISTRO, in shr.conf: # We can't set ARM_INSTRUCTION_SET_ directly since that will un-overridable in recipes like gcc Jul 13 08:00:40 PREFERRED_ARM_INSTRUCTION_SET ?= "thumb" Jul 13 08:00:40 ARM_INSTRUCTION_SET = "${PREFERRED_ARM_INSTRUCTION_SET}" Jul 13 10:17:37 Hello guys! I'm trying to use a PCI to SATA adapter in an ARM device running Yocto with an OE layer, but whenever I perform any operation on the disk filesystem libata explodes with failed commands. I would appreciate if anyone could please help me Jul 13 10:18:35 The adapter is a Silicon 3132 and the driver in use is sata_sil24 btw Jul 13 13:40:32 khem, hi Jul 13 13:40:56 one more issue wrt if.h/in.h: kexec-tools-klibc now fails with Jul 13 13:41:11 storage size of 'ifc' not known Jul 13 13:41:40 this building with musl Jul 13 13:49:00 ant_work: hmm it doesnt fail for me. Can you upload your build log somewhere Jul 13 13:50:52 Serede: perhaps you can try to get stack trace and maybe someone then could help more Jul 13 14:11:50 khem, https://filebin.net/gz1eqb3o2xrrs8r3 Jul 13 14:19:23 bbl Jul 13 15:55:38 khem, have you seen the issue? Jul 13 16:07:59 ant_home: I cant reproduce it for rpi and x86_64 Jul 13 16:08:01 on glibc Jul 13 16:08:34 so I wonder whats happening on your end might be somehow not triggered when I build for thumb Jul 13 16:08:55 just for tests can you add ARM_INSTRUCTION_SET = "thumb" to the recipe and see if that works ? Jul 13 17:12:04 ok, fiw I switched TCLIBC w/out cleaning TMPDIR Jul 13 17:14:44 same failure with ARM_INSTRUCTION_SET = "thumb" Jul 13 17:15:32 just it is built in /tmp/build/tmp-musl/work/armv5te-oe-linux-musleabi as expected Jul 13 17:15:45 that is the trigger for thumb nowadays Jul 13 17:16:57 I'll rebuild now from scratch Jul 13 17:39:56 yes Jul 13 17:40:24 ant_home: you can set the in local.conf to ensure that you use thumb ISA for every package Jul 13 17:40:42 this works great because we have already marked arm-only packages Jul 13 18:14:51 khem, what for armv4? Jul 13 18:15:23 override ìarmì ? Jul 13 18:17:12 ARM_INSTRUCTION_SET_armv4 = "arm" Jul 13 18:31:57 khem: you around? there might be an issues with glibc 2.26 and inotify on arm (at least). Where it fails with illegal instructions on a kernel without inotify support. Jul 13 18:56:48 nrossi: ok. thats not glibc issue though Jul 13 18:57:11 its probably due to the fact that we are exposing newer kernel headers as API to it Jul 13 18:57:23 nrossi: do you have some more details ? Jul 13 18:57:36 ant_home: yes Jul 13 18:58:04 khem: very likely, was seeing it on a 4.10 kernel. Just seemed odd since 2.25 was fine a couple of days ago :| Jul 13 18:58:53 oh 4.10 ? Jul 13 18:58:58 khem: i did test with musl also (same kernel/etc), which did not show the issue Jul 13 18:59:11 nrossi: then you have a solution :) Jul 13 18:59:24 on a lighter node Jul 13 18:59:32 khem: of course :D, thought i would mention it incase it comes up in other ways Jul 13 19:01:09 in anycase if you have more details on inotify issue I will help out Jul 13 19:01:37 make tests on rpi showed no regressions in inotify specifically Jul 13 19:01:46 so it must be something not covered Jul 13 19:02:55 will wait till final glibc, there seems to be some changes still going on Jul 13 19:05:51 nrossi: i would prefer now, so we can include the fic Jul 13 19:05:53 fix Jul 13 19:05:57 in final release Jul 13 19:06:47 khem: ok, will have a look at it tomorrow, need to get some sleep Jul 13 19:07:06 nrossi: ok get some msleep() Jul 13 19:26:51 khem, the build armv5e was started, kexec-tools-klibc still fails with musl Jul 13 19:27:23 now the errorlog is different Jul 13 19:30:42 khem, hm..klibc is failing, I must have pulled the old version of the patch from master Jul 13 19:31:01 for this rebuild from scratch Jul 13 22:40:43 khem, ok, it is the same error, storage size of 'ifc' isn't known Jul 13 22:41:33 ant_home: ok, it is seen on musl ? Jul 13 22:41:54 yes, from scratch, armv5e Jul 13 22:46:49 khem, I see the header in ..klibc/2.0.2-r1/recipe-sysroot/usr/include/net/if.h Jul 13 22:47:04 struct ifconf { Jul 13 22:47:04 <------>int ifc_len;<--><------> Jul 13 22:47:05 ... Jul 13 22:56:49 khem, the struct ifconf differs from glibc... Jul 13 22:58:45 the buffer is once a char pointer and once __caddr_t var? **** ENDING LOGGING AT Fri Jul 14 03:00:02 2017